Iterative active deformational methodology for tumor delineation: Evaluation across radiation treatment stage and volume
✍ Scribed by D.H. Wu; A.D. Shaffer; D.M. Thompson; Z. Yang; V.A. Magnotta; R. Alam; J. Suri; W.T.C. Yuh; N.A. Mayr
- Publisher
- John Wiley and Sons
- Year
- 2008
- Tongue
- English
- Weight
- 512 KB
- Volume
- 28
- Category
- Article
- ISSN
- 1053-1807
No coin nor oath required. For personal study only.
✦ Synopsis
Abstract
Purpose
To introduce, implement, and assess an iterative modification to the active deformational image segmentation method as applied to cervical cancer tumors.
Materials and Methods
A comparison by Jaccard similarity (J~S~) between this active deformational method and manual segmentation was performed on tumors of various sizes across preradiation, 3 weeks postradiation, and 6 weeks postradiation using a General Linear Mixed Model across 121 studies from 52 patients with Stage IIB‐IV cervical cancers.
Results
The deformable segmentation method produced promising levels of agreement including J~S~ factors of 0.71 ± 0.11 in the preradiation studies. The analysis illustrated a rate of improvement in J~S~ with increasing tumor volume that differed between the preradiation and 6 weeks postradiation stage (P = 0.0474). In the large preradiated tumors each additional cm^3^ of volume was associated with an increase or improvement in J~S~ of 0.0008 (95% confidence interval [CI]: 0.0003, 0.0014). In the smaller postradiation tumors, each additional cm^3^ of volume was associated with a more robust improvement in J~S~ of 0.0046 (95% CI: 0.0009, 0.0082).
Conclusion
Agreement was strongly affected by tumor volume, and its performance was most impacted across volume in the later stages of radiation therapy. The deformation‐based segmentation method appears to demonstrate utility for delineating cervical cancer tumors, particularly in the earliest stages of radiation treatment, where agreement is greatest. J. Magn. Reson. Imaging 2008;28:1188–1194. © 2008 Wiley‐Liss, Inc.